Over several years now I have worked sporadically to catalog hypothetical fullerenes — i.e. convex closed surfaces built of pentagons and hexagons. Last year I tossed out my old work in C to start over in Python, whose list primitive and transparent memory management made it easier to extend and generalize the project, which I now describe as enumerating and classifying roughly-convex surfaces formed of roughly-equilateral triangles. The primary goal, I suppose, is to practice my programming; another is to build a database of nonspherical “geodesic dome” shapes.
I had the idea of working the search tree in horizontal bands: find all forms with up to N vertices, saving the partial solutions for the next pass, which would read them from a file and build on them the forms with up to N+8 vertices, and so on. But my programming studies never touched on file i/o beyond the crudest, so I let that aspect sit until I learned more.
Recently I learned about Python’s shelve module, a transparent database interface which seemed just the ticket . . . until the partials file ate all the free space on my disc! <voice=”Marvin the Martian”>Back to the drawing board.</voice>